Image Enhancement. Image Enhancement

Similar documents
Digital Image Processing Chapter 3: Image Enhancement in the Spatial Domain

Image Enhancement II: Neighborhood Operations

Topic 3 - Image Enhancement. (Part 2) Spatial Filtering

Digital Image Processing

1.Discuss the frequency domain techniques of image enhancement in detail.

Image acquisition. Midterm Review. Digitization, line of image. Digitization, whole image. Geometric transformations. Interpolation 10/26/2016

CoE4TN4 Image Processing. Chapter 3: Intensity Transformation and Spatial Filtering

Digital Image Processing

Digital Image Processing

Filtering in the spatial domain (Spatial Filtering)

IMAGE PROCESSING: AREA OPERATIONS (FILTERING)

Prof. Vidya Manian Dept. of Electrical and Comptuer Engineering

Image Enhancement in spatial domain. Digital Image Processing GW Chapter 3 from Section (pag 110) Part 2: Filtering in spatial domain

Digital Image Processing

CSE 564: Visualization. Image Operations. Motivation. Provide the user (scientist, t doctor, ) with some means to: Global operations:

Practical Image and Video Processing Using MATLAB

CoE4TN4 Image Processing. Chapter 4 Filtering in the Frequency Domain

CSE 564: Scientific Visualization

EE482: Digital Signal Processing Applications

Digital Image Processing. Filtering in the Frequency Domain (Application)

Image Enhancement. DD2423 Image Analysis and Computer Vision. Computational Vision and Active Perception School of Computer Science and Communication

8.2 IMAGE PROCESSING VERSUS IMAGE ANALYSIS Image processing: The collection of routines and

Image Filtering. Median Filtering

Image Processing for feature extraction

INSTITUTE OF AERONAUTICAL ENGINEERING Dundigal, Hyderabad

LAB MANUAL SUBJECT: IMAGE PROCESSING BE (COMPUTER) SEM VII

Frequency Domain Enhancement

Vision Review: Image Processing. Course web page:

Achim J. Lilienthal Mobile Robotics and Olfaction Lab, AASS, Örebro University

DARK CURRENT ELIMINATION IN CHARGED COUPLE DEVICES

DIGITAL IMAGE DE-NOISING FILTERS A COMPREHENSIVE STUDY

Prof. Feng Liu. Winter /10/2019

Images and Filters. EE/CSE 576 Linda Shapiro

Smoothing frequency domain filters

IMAGE ENHANCEMENT IN SPATIAL DOMAIN

Transforms and Frequency Filtering

Last Lecture. photomatix.com

CSC 320 H1S CSC320 Exam Study Guide (Last updated: April 2, 2015) Winter 2015

Examples of image processing

Digital Image Processing. Digital Image Fundamentals II 12 th June, 2017

Smoothing frequency domain filters

Last Lecture. photomatix.com

TIRF, geometric operators

Table of contents. Vision industrielle 2002/2003. Local and semi-local smoothing. Linear noise filtering: example. Convolution: introduction

Midterm Review. Image Processing CSE 166 Lecture 10

Chrominance Assisted Sharpening of Images

Fourier Transforms and the Frequency Domain

Motion illusion, rotating snakes

CS/ECE 545 (Digital Image Processing) Midterm Review

Image Enhancement in the Spatial Domain

2D Signal Processing

Computing for Engineers in Python

Image Filtering Josef Pelikán & Alexander Wilkie CGG MFF UK Praha

Image filtering, image operations. Jana Kosecka

TDI2131 Digital Image Processing

IDENTIFICATION OF FISSION GAS VOIDS. Ryan Collette

Digital Image Processing

Chapter 3. Study and Analysis of Different Noise Reduction Filters

What is image enhancement? Point operation

Digital Image Processing

Templates and Image Pyramids

CS6670: Computer Vision Noah Snavely. Administrivia. Administrivia. Reading. Last time: Convolution. Last time: Cross correlation 9/8/2009

Image Enhancement using Histogram Equalization and Spatial Filtering

8. Lecture. Image restoration: Fourier domain

>>> from numpy import random as r >>> I = r.rand(256,256);

Optimized Color Transforms for Image Demosaicing

Filtering. Image Enhancement Spatial and Frequency Based

Templates and Image Pyramids

24 Bit Image Noise Reduction with Median Filtering Algorithm

VU Signal and Image Processing. Image Enhancement. Torsten Möller + Hrvoje Bogunović + Raphael Sahann

Anna University, Chennai B.E./B.TECH DEGREE EXAMINATION, MAY/JUNE 2013 Seventh Semester

Filip Malmberg 1TD396 fall 2018 Today s lecture

Numerical Derivatives See also T&V, Appendix A.2 Gradient = vector of partial derivatives of image I(x,y) = [di(x,y)/dx, di(x,y)/dy]

International Journal of Computer Engineering and Applications, Volume XI, Issue IX, September 17, ISSN

CS 4501: Introduction to Computer Vision. Filtering and Edge Detection

Digital Image Processing. Image Enhancement: Filtering in the Frequency Domain

Spatial Domain Processing and Image Enhancement

Chapter 3 Image Enhancement in the Spatial Domain. Chapter 3 Image Enhancement in the Spatial Domain

PRACTICAL IMAGE AND VIDEO PROCESSING USING MATLAB

DIGITAL IMAGE PROCESSING ASSIGNMENT

Digital Image Processing Chapter 6: Color Image Processing

An Efficient Noise Removing Technique Using Mdbut Filter in Images

Fourier Transform. Any signal can be expressed as a linear combination of a bunch of sine gratings of different frequency Amplitude Phase

June 30 th, 2008 Lesson notes taken from professor Hongmei Zhu class.

Computer Vision, Lecture 3

Digital Image Processing 3/e

Chapter 2 Image Enhancement in the Spatial Domain

Enhancement. Degradation model H and noise must be known/predicted first before restoration. Noise model Degradation Model

Motivation: Image denoising. How can we reduce noise in a photograph?

Introduction Approach Work Performed and Results

Circular averaging filter (pillbox) Approximates the two-dimensional Laplacian operator. Laplacian of Gaussian filter

Image Processing. Adam Finkelstein Princeton University COS 426, Spring 2019

Noise and Restoration of Images

ECE5984 Orthogonal Frequency Division Multiplexing and Related Technologies Fall Mohamed Essam Khedr. Channel Estimation

Image Enhancement in the Spatial Domain Low and High Pass Filtering

Motivation: Image denoising. How can we reduce noise in a photograph?

Area Extraction of beads in Membrane filter using Image Segmentation Techniques

1. (a) Explain the process of Image acquisition. (b) Discuss different elements used in digital image processing system. [8+8]

Head, IICT, Indus University, India

Image Filtering in Spatial domain. Computer Vision Jia-Bin Huang, Virginia Tech

Transcription:

SPATIAL FILTERING g h * h g FREQUENCY DOMAIN FILTERING G H. F F H G Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens LOW PASS FILTERING attenuate or eliminate high-requenc components edges and other sharp details results in image bluring HIGH PASS FILTERING attenuate or eliminate low-requenc components slowl varing characteristics such as overall contrast and average intensit results in reduction o overall contrast and average intensit and a correspondingl apparent sharpening o edges and other sharp details Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens

Frequenc domain ilters G H. F Spatial domain ilters g h * Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 3 SMOOTHING FILTERS - LOWPASS SPATIAL FILTERING With the assumption that image piel values within a small neighborhood are highl correlated and that the noise components are not correlated noise ma be reduced b replacing each piel with the mean over a certain neighborhood o : g n m µ. M n m S where M is the number o piels in the neighborhood S. This is useul when onl one version o the image is available. I this operation is perormed over a 3 3 neighborhood we have g 9 i j + i + Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 4 j.

CONVOLUTION BY MASK OPERATION: The 33 mean ilter ma be epressed b the convolution mask 9 Unortunatel the mean ilter operation blurs edges and sharp eatures. g h * Convolution mask or kernel Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 5 Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 6 3

a Original image; b- results o spatial lowpass iltering with masks size o 33 55 77 55 55. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 7 Blurring o edges ma be controlled b selective mean iltering: µ i µ > T g otherwise. where T is a threshold this is useul in salt and pepper noise in this applications the central piel at is usuall let out to use onl the eight neighboring piel in computing the mean. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 4

SMOOTHING FILTERS - MEDIAN FILTERING non-linear ilter perorms better noise removal with less blurring in most cases. 5 5 5 5 5 5 sorting: 5 5 median Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 9 a Original image; b image corrupted b impulse noise; c result o 55 mean; d result o 55 median iltering. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 5

6 Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens SHARPENING FILTERS - HIGHPASS SPATIAL FILTERING Edge Enhancement and Etraction The gradient operator gives a measure o change in the image values in the direction speciied: For digital dierentiation is approimated b dierences: j i G +. + G [ ] [ ]. + + G or G Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens Dierentiation leads to removal o constant values in the direction o the operation; etraction o edges in the orthogonal direction; and removal o the average intensit DC component. Roberts gradient uses cross-dierences This operator computes diagonal edge gradients.the advantage o this operator is that the resulting image piel values ma be written in the same arra as the input image. [ ] [ ]. G + + + + +

7 Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 3 3 3 MASKS FOR GRADIENT OPERATIONS Prewitt operators: Sobel operators: 3 ; 3 4 ; 4 Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 4 33 MASK FOR IMAGE SHARPENING Laplacian: Subtracting Laplacian: Unsharp Masking: 4 / / / / / / / / 5

Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 5 3X3 MASK FOR DIRECTIONAL GRADIENTS. : 9 ; : o o. : 35 ; : 45 o o Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 6 EXAMPLES OF 3X3 MASK OPERATIONS: 4 * 3 3 3 3 5 *

9 Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 7 9 Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 9 w Where w 9A - with A a original image; b A.; c A.5; d A..

a original image; b magnitude o Prewitt gradient; c setting to 55 an gradient value over 5; d setting to 55 an gradient value over 5 and setting to an gradient value under or equal 5. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 9 a original image; b vertical edge detector; c horizontal edge detector; d Sobel edge detector; e Roberts edge detector. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens

FREQUENCY DOMAIN FILTERING High-requenc components are associated with sharp eatures in the image as well as noise. To achieve smoothing o images and/or noise removal we ma remove or attenuate a certain portion o the highrequenc components b lowpass iltering. G H. F F H G Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens LOWPASS FILTER FUNCTIONS: ideal: H i D D o otherwise Note : D u + v i. e. the radial requenc. G H. F Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens

a 55 image; b its Fourier spectrum with superimposed circles which radii equal to 43 7 and 5 enclose 9 93 95 99 and 99.5% o the image power respectivel. P T N N u v P P F Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 3 a original image; b- results o ideal lowpass iltering with the cuto requenc set at the radii equal to 43 7 and 5 respectivel. Ideal losspass ilters results in blurring and ringing removing edge and sharp detail inormation o the image Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 4

h G H. F g h * Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 5 While "ideal" iltering is possible on computers it is not desirable as it results in ringing artiacts around edges in the image. Eponential and Butterworth ilters provide a smoother roll o and produce smooth images with no ringing artiacts. Eponential: D H ep D o n. D + u v Butterworth: H u D u + D o n. Note: n is the order o the ilter; higher-order ilters provide aster roll-o. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 6 3

a original image; b- results o Butterworth lowpass iltering with the cuto requenc set at the radii equal to 43 7 and 5 respectivel. Less blurring and no ringing Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 7 a image digitized with onl 6 gra levels ehibits alse contours; b result o smothing a with a lowpass ilter o order ; c nois image; d results o appling Butterworth lowpass iltering to the nois image. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 4

HIGHPASS FILTER FUNCTIONS: Highpass ilters are useul in edge etraction applications. Ideal: H i D D otherwise. o Eponential: n D o H ep. D u v Butterworth: H. n Do + D Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 9 a original image; b result ater a highpass Butterworth ilter > lowrequenc components were severel attenuated thus making dierent gra-level regions appear the same c result ater high-requenc emphasis high-requenc emphasis ilter highpass ilter + a constant; d results o appling highrequenc emphasis and histogram equalization. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 3 5

Directional "sector" ilters ma be designed to enhance etract or remove eatures at preerred orientations b virtue o the rotational propert o the Fourier transorm. While space domain operations aect local piel values and eatures requenc domain operations aect the image globall. While normall we are concerned with the magnitude spectrum to a large etent the phase spectrum is also important. Phase has been shown to be associated with edge inormation to a larger etent than the magnitude o the requenc components. Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 3 HOMOMORPHIC FILTERING i. r F I. R i: illumination component ver low requenc; r: relectance component medium-to-high requenc. To separate the two components or iltering take the logarithm: [. ] ln[ i ] ln[ r ] z ln + Z I + R u. S H. Z [ s ] g ep ln FFT H FFT - ep g Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 3 6

a original image; b image processed b homomorphic iltering to achieve simultaneous dnamic range compression and contrast enhancement. b enhancing r and suppressing i Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 33 Color Image Processing Slide projector Copright RMR / RDL - 999. PEE53 - Processamento Digital de Imagens 34 7